1 /* $OpenBSD: kroute.c,v 1.34 2019/12/11 21:04:59 remi Exp $ */ 2 3 /* 4 * Copyright (c) 2004 Esben Norby <norby@openbsd.org> 5 * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org> 6 * 7 * Permission to use, copy, modify, and distribute this software for any 8 * purpose with or without fee is hereby granted, provided that the above 9 * copyright notice and this permission notice appear in all copies. 10 * 11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 18 */ 19 20 #include <sys/types.h> 21 #include <sys/socket.h> 22 #include <sys/sysctl.h> 23 #include <sys/tree.h> 24 #include <sys/uio.h> 25 #include <netinet/in.h> 26 #include <arpa/inet.h> 27 #include <net/if.h> 28 #include <net/if_dl.h> 29 #include <net/if_types.h> 30 #include <net/route.h> 31 #include <err.h> 32 #include <errno.h> 33 #include <fcntl.h> 34 #include <stdio.h> 35 #include <stdlib.h> 36 #include <string.h> 37 #include <unistd.h> 38 39 #include "rip.h" 40 #include "ripd.h" 41 #include "log.h" 42 43 struct { 44 u_int32_t rtseq; 45 pid_t pid; 46 int fib_sync; 47 u_int8_t fib_prio; 48 int fd; 49 struct event ev; 50 u_int rdomain; 51 } kr_state; 52 53 struct kroute_node { 54 RB_ENTRY(kroute_node) entry; 55 struct kroute r; 56 }; 57 58 struct kif_node { 59 RB_ENTRY(kif_node) entry; 60 struct kif k; 61 }; 62 63 void kr_redistribute(int, struct kroute *); 64 int kroute_compare(struct kroute_node *, struct kroute_node *); 65 int kif_compare(struct kif_node *, struct kif_node *); 66 int kr_change_fib(struct kroute_node *, struct kroute *, int); 67 68 struct kroute_node *kroute_find(in_addr_t, in_addr_t, u_int8_t); 69 int kroute_insert(struct kroute_node *); 70 int kroute_remove(struct kroute_node *); 71 void kroute_clear(void); 72 73 struct kif_node *kif_find(int); 74 int kif_insert(struct kif_node *); 75 int kif_remove(struct kif_node *); 76 void kif_clear(void); 77 int kif_validate(int); 78 79 struct kroute_node *kroute_match(in_addr_t); 80 81 int protect_lo(void); 82 u_int8_t prefixlen_classful(in_addr_t); 83 void get_rtaddrs(int, struct sockaddr *, struct sockaddr **); 84 void if_change(u_short, int, struct if_data *); 85 void if_announce(void *); 86 87 int send_rtmsg(int, int, struct kroute *); 88 int dispatch_rtmsg(void); 89 int fetchtable(void); 90 int fetchifs(int); 91 92 RB_HEAD(kroute_tree, kroute_node) krt; 93 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare) 94 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare) 95 96 RB_HEAD(kif_tree, kif_node) kit; 97 RB_PROTOTYPE(kif_tree, kif_node, entry, kif_compare) 98 RB_GENERATE(kif_tree, kif_node, entry, kif_compare) 99 100 int 101 kif_init(void) 102 { 103 RB_INIT(&kit); 104 105 if (fetchifs(0) == -1) 106 return (-1); 107 108 return (0); 109 } 110 111 int 112 kr_init(int fs, u_int rdomain, u_int8_t fib_prio) 113 { 114 int opt = 0, rcvbuf, default_rcvbuf; 115 socklen_t optlen; 116 117 if ((kr_state.fd = socket(AF_ROUTE, 118 SOCK_RAW | SOCK_CLOEXEC | SOCK_NONBLOCK, 0)) == -1) { 119 log_warn("kr_init: socket"); 120 return (-1); 121 } 122 123 /* not interested in my own messages */ 124 if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK, 125 &opt, sizeof(opt)) == -1) 126 log_warn("kr_init: setsockopt"); /* not fatal */ 127 128 /* grow receive buffer, don't wanna miss messages */ 129 optlen = sizeof(default_rcvbuf); 130 if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, 131 &default_rcvbuf, &optlen) == -1) 132 log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF"); 133 else 134 for (rcvbuf = MAX_RTSOCK_BUF; 135 rcvbuf > default_rcvbuf && 136 setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, 137 &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS; 138 rcvbuf /= 2) 139 ; /* nothing */ 140 141 kr_state.pid = getpid(); 142 kr_state.rtseq = 1; 143 kr_state.fib_prio = fib_prio; 144 145 RB_INIT(&krt); 146 147 if (fetchtable() == -1) 148 return (-1); 149 150 if (protect_lo() == -1) 151 return (-1); 152 153 kr_state.fib_sync = fs; /* now set correct sync mode */ 154 kr_state.rdomain = rdomain; 155 156 event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST, 157 kr_dispatch_msg, NULL); 158 event_add(&kr_state.ev, NULL); 159 160 return (0); 161 } 162 163 int 164 kr_change_fib(struct kroute_node *kr, struct kroute *kroute, int action) 165 { 166 /* nexthop within 127/8 -> ignore silently */ 167 if ((kroute->nexthop.s_addr & htonl(IN_CLASSA_NET)) == 168 htonl(INADDR_LOOPBACK & IN_CLASSA_NET)) 169 return (0); 170 171 if (send_rtmsg(kr_state.fd, action, kroute) == -1) 172 return (-1); 173 174 if (action == RTM_ADD) { 175 if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) 176 fatal("kr_change_fib"); 177 178 kr->r.prefix.s_addr = kroute->prefix.s_addr; 179 kr->r.netmask.s_addr = kroute->netmask.s_addr; 180 kr->r.nexthop.s_addr = kroute->nexthop.s_addr; 181 kr->r.flags = kroute->flags |= F_RIPD_INSERTED; 182 kr->r.priority = kr_state.fib_prio; 183 184 if (kroute_insert(kr) == -1) { 185 log_debug("kr_update_fib: cannot insert %s", 186 inet_ntoa(kroute->nexthop)); 187 } 188 } else 189 kr->r.nexthop.s_addr = kroute->nexthop.s_addr; 190 191 return (0); 192 } 193 194 int 195 kr_change(struct kroute *kroute) 196 { 197 struct kroute_node *kr; 198 int action = RTM_ADD; 199 200 kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr, 201 kr_state.fib_prio); 202 if (kr != NULL) 203 action = RTM_CHANGE; 204 205 return (kr_change_fib(kr, kroute, action)); 206 } 207 208 int 209 kr_delete(struct kroute *kroute) 210 { 211 struct kroute_node *kr; 212 213 kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr, 214 kr_state.fib_prio); 215 if (kr == NULL) 216 return (0); 217 218 if (kr->r.priority != kr_state.fib_prio) 219 log_warn("kr_delete_fib: %s/%d has wrong priority %d", 220 inet_ntoa(kr->r.prefix), mask2prefixlen(kr->r.netmask.s_addr), 221 kr->r.priority); 222 223 if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1) 224 return (-1); 225 226 if (kroute_remove(kr) == -1) 227 return (-1); 228 229 return (0); 230 } 231 232 void 233 kr_shutdown(void) 234 { 235 kr_fib_decouple(); 236 237 kroute_clear(); 238 kif_clear(); 239 } 240 241 void 242 kr_fib_couple(void) 243 { 244 struct kroute_node *kr; 245 246 if (kr_state.fib_sync == 1) /* already coupled */ 247 return; 248 249 kr_state.fib_sync = 1; 250 251 RB_FOREACH(kr, kroute_tree, &krt) 252 if (kr->r.priority == kr_state.fib_prio) 253 send_rtmsg(kr_state.fd, RTM_ADD, &kr->r); 254 255 log_info("kernel routing table coupled"); 256 } 257 258 void 259 kr_fib_decouple(void) 260 { 261 struct kroute_node *kr; 262 263 if (kr_state.fib_sync == 0) /* already decoupled */ 264 return; 265 266 RB_FOREACH(kr, kroute_tree, &krt) 267 if (kr->r.priority == kr_state.fib_prio) 268 send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r); 269 270 kr_state.fib_sync = 0; 271 272 log_info("kernel routing table decoupled"); 273 } 274 275 /* ARGSUSED */ 276 void 277 kr_dispatch_msg(int fd, short event, void *bula) 278 { 279 dispatch_rtmsg(); 280 } 281 282 void 283 kr_show_route(struct imsg *imsg) 284 { 285 struct kroute_node *kr; 286 int flags; 287 struct in_addr addr; 288 289 switch (imsg->hdr.type) { 290 case IMSG_CTL_KROUTE: 291 if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) { 292 log_warnx("kr_show_route: wrong imsg len"); 293 return; 294 } 295 memcpy(&flags, imsg->data, sizeof(flags)); 296 RB_FOREACH(kr, kroute_tree, &krt) 297 if (!flags || kr->r.flags & flags) { 298 main_imsg_compose_ripe(IMSG_CTL_KROUTE, 299 imsg->hdr.pid, &kr->r, sizeof(kr->r)); 300 } 301 break; 302 case IMSG_CTL_KROUTE_ADDR: 303 if (imsg->hdr.len != IMSG_HEADER_SIZE + 304 sizeof(struct in_addr)) { 305 log_warnx("kr_show_route: wrong imsg len"); 306 return; 307 } 308 memcpy(&addr, imsg->data, sizeof(addr)); 309 kr = NULL; 310 kr = kroute_match(addr.s_addr); 311 if (kr != NULL) 312 main_imsg_compose_ripe(IMSG_CTL_KROUTE, imsg->hdr.pid, 313 &kr->r, sizeof(kr->r)); 314 break; 315 default: 316 log_debug("kr_show_route: error handling imsg"); 317 break; 318 } 319 320 main_imsg_compose_ripe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0); 321 } 322 323 void 324 kr_ifinfo(char *ifname, pid_t pid) 325 { 326 struct kif_node *kif; 327 328 RB_FOREACH(kif, kif_tree, &kit) 329 if (ifname == NULL || !strcmp(ifname, kif->k.ifname)) { 330 main_imsg_compose_ripe(IMSG_CTL_IFINFO, 331 pid, &kif->k, sizeof(kif->k)); 332 } 333 334 main_imsg_compose_ripe(IMSG_CTL_END, pid, NULL, 0); 335 } 336 337 void 338 kr_redistribute(int type, struct kroute *kr) 339 { 340 u_int32_t a; 341 342 343 if (type == IMSG_NETWORK_DEL) { 344 dont_redistribute: 345 /* was the route redistributed? */ 346 if (kr->flags & F_REDISTRIBUTED) { 347 /* remove redistributed flag */ 348 kr->flags &= ~F_REDISTRIBUTED; 349 main_imsg_compose_rde(type, 0, kr, 350 sizeof(struct kroute)); 351 } 352 return; 353 } 354 355 /* interface is not up and running so don't announce */ 356 if (kr->flags & F_DOWN) 357 return; 358 359 /* 360 * We consider the loopback net, multicast and experimental addresses 361 * as not redistributable. 362 */ 363 a = ntohl(kr->prefix.s_addr); 364 if (IN_MULTICAST(a) || IN_BADCLASS(a) || 365 (a >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 366 return; 367 /* 368 * Consider networks with nexthop loopback as not redistributable 369 * unless it is a reject or blackhole route. 370 */ 371 if (kr->nexthop.s_addr == htonl(INADDR_LOOPBACK) && 372 !(kr->flags & (F_BLACKHOLE|F_REJECT))) 373 return; 374 375 /* Should we redistribute this route? */ 376 if (!rip_redistribute(kr)) 377 goto dont_redistribute; 378 379 /* Does not matter if we resend the kr, the RDE will cope. */ 380 kr->flags |= F_REDISTRIBUTED; 381 main_imsg_compose_rde(type, 0, kr, sizeof(struct kroute)); 382 } 383 384 /* rb-tree compare */ 385 int 386 kroute_compare(struct kroute_node *a, struct kroute_node *b) 387 { 388 if (ntohl(a->r.prefix.s_addr) < ntohl(b->r.prefix.s_addr)) 389 return (-1); 390 if (ntohl(a->r.prefix.s_addr) > ntohl(b->r.prefix.s_addr)) 391 return (1); 392 if (ntohl(a->r.netmask.s_addr) < ntohl(b->r.netmask.s_addr)) 393 return (-1); 394 if (ntohl(a->r.netmask.s_addr) > ntohl(b->r.netmask.s_addr)) 395 return (1); 396 397 /* if the priority is RTP_ANY finish on the first address hit */ 398 if (a->r.priority == RTP_ANY || b->r.priority == RTP_ANY) 399 return (0); 400 if (a->r.priority < b->r.priority) 401 return (-1); 402 if (a->r.priority > b->r.priority) 403 return (1); 404 405 return (0); 406 } 407 408 int 409 kif_compare(struct kif_node *a, struct kif_node *b) 410 { 411 return (b->k.ifindex - a->k.ifindex); 412 } 413 414 /* tree management */ 415 struct kroute_node * 416 kroute_find(in_addr_t prefix, in_addr_t netmask, u_int8_t prio) 417 { 418 struct kroute_node s, *kn, *tmp; 419 420 s.r.prefix.s_addr = prefix; 421 s.r.netmask.s_addr = netmask; 422 s.r.priority = prio; 423 424 kn = RB_FIND(kroute_tree, &krt, &s); 425 if (kn && prio == RTP_ANY) { 426 tmp = RB_PREV(kroute_tree, &krt, kn); 427 while (tmp) { 428 if (kroute_compare(&s, tmp) == 0) 429 kn = tmp; 430 else 431 break; 432 tmp = RB_PREV(kroute_tree, &krt, kn); 433 } 434 } 435 436 return (kn); 437 } 438 439 int 440 kroute_insert(struct kroute_node *kr) 441 { 442 if (RB_INSERT(kroute_tree, &krt, kr) != NULL) { 443 log_warnx("kroute_insert failed for %s/%u", 444 inet_ntoa(kr->r.prefix), 445 mask2prefixlen(kr->r.netmask.s_addr)); 446 free(kr); 447 return (-1); 448 } 449 450 if (!(kr->r.flags & F_KERNEL)) { 451 /* don't validate or redistribute rip route */ 452 kr->r.flags &= ~F_DOWN; 453 return (0); 454 } 455 456 if (kif_validate(kr->r.ifindex)) 457 kr->r.flags &= ~F_DOWN; 458 else 459 kr->r.flags |= F_DOWN; 460 461 kr_redistribute(IMSG_NETWORK_ADD, &kr->r); 462 463 return (0); 464 } 465 466 int 467 kroute_remove(struct kroute_node *kr) 468 { 469 if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) { 470 log_warnx("kroute_remove failed for %s/%u", 471 inet_ntoa(kr->r.prefix), 472 mask2prefixlen(kr->r.netmask.s_addr)); 473 return (-1); 474 } 475 476 kr_redistribute(IMSG_NETWORK_DEL, &kr->r); 477 rtlabel_unref(kr->r.rtlabel); 478 479 free(kr); 480 return (0); 481 } 482 483 void 484 kroute_clear(void) 485 { 486 struct kroute_node *kr; 487 488 while ((kr = RB_MIN(kroute_tree, &krt)) != NULL) 489 kroute_remove(kr); 490 } 491 492 struct kif_node * 493 kif_find(int ifindex) 494 { 495 struct kif_node s; 496 497 bzero(&s, sizeof(s)); 498 s.k.ifindex = ifindex; 499 500 return (RB_FIND(kif_tree, &kit, &s)); 501 } 502 503 struct kif * 504 kif_findname(char *ifname) 505 { 506 struct kif_node *kif; 507 508 RB_FOREACH(kif, kif_tree, &kit) 509 if (!strcmp(ifname, kif->k.ifname)) 510 return (&kif->k); 511 512 return (NULL); 513 } 514 515 int 516 kif_insert(struct kif_node *kif) 517 { 518 if (RB_INSERT(kif_tree, &kit, kif) != NULL) { 519 log_warnx("RB_INSERT(kif_tree, &kit, kif)"); 520 free(kif); 521 return (-1); 522 } 523 524 return (0); 525 } 526 527 int 528 kif_remove(struct kif_node *kif) 529 { 530 if (RB_REMOVE(kif_tree, &kit, kif) == NULL) { 531 log_warnx("RB_REMOVE(kif_tree, &kit, kif)"); 532 return (-1); 533 } 534 535 free(kif); 536 return (0); 537 } 538 539 void 540 kif_clear(void) 541 { 542 struct kif_node *kif; 543 544 while ((kif = RB_MIN(kif_tree, &kit)) != NULL) 545 kif_remove(kif); 546 } 547 548 int 549 kif_validate(int ifindex) 550 { 551 struct kif_node *kif; 552 553 if ((kif = kif_find(ifindex)) == NULL) { 554 log_warnx("interface with index %u not found", ifindex); 555 return (1); 556 } 557 558 return (kif->k.nh_reachable); 559 } 560 561 struct kroute_node * 562 kroute_match(in_addr_t key) 563 { 564 u_int8_t i; 565 struct kroute_node *kr; 566 567 /* we will never match the default route */ 568 for (i = 32; i > 0; i--) 569 if ((kr = kroute_find(key & prefixlen2mask(i), 570 prefixlen2mask(i), RTP_ANY)) != NULL) 571 return (kr); 572 573 /* if we don't have a match yet, try to find a default route */ 574 if ((kr = kroute_find(0, 0, RTP_ANY)) != NULL) 575 return (kr); 576 577 return (NULL); 578 } 579 580 /* misc */ 581 int 582 protect_lo(void) 583 { 584 struct kroute_node *kr; 585 586 /* special protection for 127/8 */ 587 if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) { 588 log_warn("protect_lo"); 589 return (-1); 590 } 591 kr->r.prefix.s_addr = htonl(INADDR_LOOPBACK); 592 kr->r.netmask.s_addr = htonl(IN_CLASSA_NET); 593 kr->r.flags = F_KERNEL|F_CONNECTED; 594 595 if (RB_INSERT(kroute_tree, &krt, kr) != NULL) 596 free(kr); /* kernel route already there, no problem */ 597 598 return (0); 599 } 600 601 u_int8_t 602 prefixlen_classful(in_addr_t ina) 603 { 604 /* it hurt to write this. */ 605 606 if (ina >= 0xf0000000U) /* class E */ 607 return (32); 608 else if (ina >= 0xe0000000U) /* class D */ 609 return (4); 610 else if (ina >= 0xc0000000U) /* class C */ 611 return (24); 612 else if (ina >= 0x80000000U) /* class B */ 613 return (16); 614 else /* class A */ 615 return (8); 616 } 617 618 u_int8_t 619 mask2prefixlen(in_addr_t ina) 620 { 621 if (ina == 0) 622 return (0); 623 else 624 return (33 - ffs(ntohl(ina))); 625 } 626 627 in_addr_t 628 prefixlen2mask(u_int8_t prefixlen) 629 { 630 if (prefixlen == 0) 631 return (0); 632 633 return (htonl(0xffffffff << (32 - prefixlen))); 634 } 635 636 #define ROUNDUP(a) \ 637 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 638 639 void 640 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info) 641 { 642 int i; 643 644 for (i = 0; i < RTAX_MAX; i++) { 645 if (addrs & (1 << i)) { 646 rti_info[i] = sa; 647 sa = (struct sockaddr *)((char *)(sa) + 648 ROUNDUP(sa->sa_len)); 649 } else 650 rti_info[i] = NULL; 651 } 652 } 653 654 void 655 if_change(u_short ifindex, int flags, struct if_data *ifd) 656 { 657 struct kif_node *kif; 658 struct kroute_node *kr; 659 int type; 660 u_int8_t reachable; 661 662 if ((kif = kif_find(ifindex)) == NULL) { 663 log_warnx("interface with index %u not found", ifindex); 664 return; 665 } 666 667 kif->k.flags = flags; 668 kif->k.link_state = ifd->ifi_link_state; 669 kif->k.if_type = ifd->ifi_type; 670 kif->k.baudrate = ifd->ifi_baudrate; 671 672 if ((reachable = (flags & IFF_UP) && 673 LINK_STATE_IS_UP(ifd->ifi_link_state)) == kif->k.nh_reachable) 674 return; /* nothing changed wrt nexthop validity */ 675 676 kif->k.nh_reachable = reachable; 677 type = reachable ? IMSG_NETWORK_ADD : IMSG_NETWORK_DEL; 678 679 /* notify ripe about interface link state */ 680 main_imsg_compose_ripe(IMSG_IFINFO, 0, &kif->k, sizeof(kif->k)); 681 682 /* update redistribute list */ 683 RB_FOREACH(kr, kroute_tree, &krt) 684 if (kr->r.ifindex == ifindex) { 685 if (reachable) 686 kr->r.flags &= ~F_DOWN; 687 else 688 kr->r.flags |= F_DOWN; 689 690 kr_redistribute(type, &kr->r); 691 } 692 } 693 694 void 695 if_announce(void *msg) 696 { 697 struct if_announcemsghdr *ifan; 698 struct kif_node *kif; 699 700 ifan = msg; 701 702 switch (ifan->ifan_what) { 703 case IFAN_ARRIVAL: 704 if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) { 705 log_warn("if_announce"); 706 return; 707 } 708 709 kif->k.ifindex = ifan->ifan_index; 710 strlcpy(kif->k.ifname, ifan->ifan_name, sizeof(kif->k.ifname)); 711 kif_insert(kif); 712 break; 713 case IFAN_DEPARTURE: 714 kif = kif_find(ifan->ifan_index); 715 kif_remove(kif); 716 break; 717 } 718 } 719 720 /* rtsock */ 721 int 722 send_rtmsg(int fd, int action, struct kroute *kroute) 723 { 724 struct iovec iov[4]; 725 struct rt_msghdr hdr; 726 struct sockaddr_in prefix; 727 struct sockaddr_in nexthop; 728 struct sockaddr_in mask; 729 int iovcnt = 0; 730 731 if (kr_state.fib_sync == 0) 732 return (0); 733 734 /* initialize header */ 735 bzero(&hdr, sizeof(hdr)); 736 hdr.rtm_version = RTM_VERSION; 737 hdr.rtm_type = action; 738 hdr.rtm_priority = kr_state.fib_prio; 739 hdr.rtm_tableid = kr_state.rdomain; 740 if (action == RTM_CHANGE) 741 hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE; 742 hdr.rtm_seq = kr_state.rtseq++; /* overflow doesn't matter */ 743 hdr.rtm_msglen = sizeof(hdr); 744 /* adjust iovec */ 745 iov[iovcnt].iov_base = &hdr; 746 iov[iovcnt++].iov_len = sizeof(hdr); 747 748 bzero(&prefix, sizeof(prefix)); 749 prefix.sin_len = sizeof(prefix); 750 prefix.sin_family = AF_INET; 751 prefix.sin_addr.s_addr = kroute->prefix.s_addr; 752 /* adjust header */ 753 hdr.rtm_addrs |= RTA_DST; 754 hdr.rtm_msglen += sizeof(prefix); 755 /* adjust iovec */ 756 iov[iovcnt].iov_base = &prefix; 757 iov[iovcnt++].iov_len = sizeof(prefix); 758 759 if (kroute->nexthop.s_addr != 0) { 760 bzero(&nexthop, sizeof(nexthop)); 761 nexthop.sin_len = sizeof(nexthop); 762 nexthop.sin_family = AF_INET; 763 nexthop.sin_addr.s_addr = kroute->nexthop.s_addr; 764 /* adjust header */ 765 hdr.rtm_flags |= RTF_GATEWAY; 766 hdr.rtm_addrs |= RTA_GATEWAY; 767 hdr.rtm_msglen += sizeof(nexthop); 768 /* adjust iovec */ 769 iov[iovcnt].iov_base = &nexthop; 770 iov[iovcnt++].iov_len = sizeof(nexthop); 771 } 772 773 bzero(&mask, sizeof(mask)); 774 mask.sin_len = sizeof(mask); 775 mask.sin_family = AF_INET; 776 mask.sin_addr.s_addr = kroute->netmask.s_addr; 777 /* adjust header */ 778 hdr.rtm_addrs |= RTA_NETMASK; 779 hdr.rtm_msglen += sizeof(mask); 780 /* adjust iovec */ 781 iov[iovcnt].iov_base = &mask; 782 iov[iovcnt++].iov_len = sizeof(mask); 783 784 785 retry: 786 if (writev(fd, iov, iovcnt) == -1) { 787 if (errno == ESRCH) { 788 if (hdr.rtm_type == RTM_CHANGE) { 789 hdr.rtm_type = RTM_ADD; 790 goto retry; 791 } else if (hdr.rtm_type == RTM_DELETE) { 792 log_info("route %s/%u vanished before delete", 793 inet_ntoa(kroute->prefix), 794 mask2prefixlen(kroute->netmask.s_addr)); 795 return (0); 796 } 797 } 798 log_warn("send_rtmsg: action %u, prefix %s/%u", 799 hdr.rtm_type, inet_ntoa(kroute->prefix), 800 mask2prefixlen(kroute->netmask.s_addr)); 801 return (0); 802 } 803 804 return (0); 805 } 806 807 int 808 fetchtable(void) 809 { 810 size_t len; 811 int mib[7]; 812 char *buf, *next, *lim; 813 struct rt_msghdr *rtm; 814 struct sockaddr *sa, *rti_info[RTAX_MAX]; 815 struct sockaddr_in *sa_in; 816 struct sockaddr_rtlabel *label; 817 struct kroute_node *kr; 818 struct iface *iface = NULL; 819 820 mib[0] = CTL_NET; 821 mib[1] = PF_ROUTE; 822 mib[2] = 0; 823 mib[3] = AF_INET; 824 mib[4] = NET_RT_DUMP; 825 mib[5] = 0; 826 mib[6] = kr_state.rdomain; /* rtableid */ 827 828 if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) { 829 log_warn("sysctl"); 830 return (-1); 831 } 832 if ((buf = malloc(len)) == NULL) { 833 log_warn("fetchtable"); 834 return (-1); 835 } 836 if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) { 837 log_warn("sysctl"); 838 free(buf); 839 return (-1); 840 } 841 842 lim = buf + len; 843 for (next = buf; next < lim; next += rtm->rtm_msglen) { 844 rtm = (struct rt_msghdr *)next; 845 if (rtm->rtm_version != RTM_VERSION) 846 continue; 847 sa = (struct sockaddr *)(next + rtm->rtm_hdrlen); 848 get_rtaddrs(rtm->rtm_addrs, sa, rti_info); 849 850 if ((sa = rti_info[RTAX_DST]) == NULL) 851 continue; 852 853 /* Skip ARP/ND cache and broadcast routes. */ 854 if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST)) 855 continue; 856 857 #ifdef RTF_MPATH 858 if (rtm->rtm_flags & RTF_MPATH) /* multipath */ 859 continue; 860 #endif 861 862 if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) { 863 log_warn("fetchtable"); 864 free(buf); 865 return (-1); 866 } 867 868 kr->r.flags = F_KERNEL; 869 kr->r.priority = rtm->rtm_priority; 870 871 switch (sa->sa_family) { 872 case AF_INET: 873 kr->r.prefix.s_addr = 874 ((struct sockaddr_in *)sa)->sin_addr.s_addr; 875 sa_in = (struct sockaddr_in *)rti_info[RTAX_NETMASK]; 876 if (rtm->rtm_flags & RTF_STATIC) 877 kr->r.flags |= F_STATIC; 878 if (rtm->rtm_flags & RTF_BLACKHOLE) 879 kr->r.flags |= F_BLACKHOLE; 880 if (rtm->rtm_flags & RTF_REJECT) 881 kr->r.flags |= F_REJECT; 882 if (rtm->rtm_flags & RTF_DYNAMIC) 883 kr->r.flags |= F_DYNAMIC; 884 if (sa_in != NULL) { 885 if (sa_in->sin_len == 0) 886 break; 887 kr->r.netmask.s_addr = 888 sa_in->sin_addr.s_addr; 889 } else if (rtm->rtm_flags & RTF_HOST) 890 kr->r.netmask.s_addr = prefixlen2mask(32); 891 else 892 kr->r.netmask.s_addr = 893 prefixlen2mask(prefixlen_classful 894 (kr->r.prefix.s_addr)); 895 break; 896 default: 897 free(kr); 898 continue; 899 } 900 901 kr->r.ifindex = rtm->rtm_index; 902 903 iface = if_find_index(rtm->rtm_index); 904 if (iface != NULL) 905 kr->r.metric = iface->cost; 906 else 907 kr->r.metric = DEFAULT_COST; 908 909 if ((sa = rti_info[RTAX_GATEWAY]) != NULL) 910 switch (sa->sa_family) { 911 case AF_INET: 912 if (rtm->rtm_flags & RTF_CONNECTED) { 913 kr->r.flags |= F_CONNECTED; 914 break; 915 } 916 917 kr->r.nexthop.s_addr = 918 ((struct sockaddr_in *)sa)->sin_addr.s_addr; 919 break; 920 case AF_LINK: 921 /* 922 * Traditional BSD connected routes have 923 * a gateway of type AF_LINK. 924 */ 925 kr->r.flags |= F_CONNECTED; 926 break; 927 } 928 929 if (rtm->rtm_priority == kr_state.fib_prio) { 930 send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r); 931 free(kr); 932 } else { 933 if ((label = (struct sockaddr_rtlabel *) 934 rti_info[RTAX_LABEL]) != NULL) 935 kr->r.rtlabel = 936 rtlabel_name2id(label->sr_label); 937 kroute_insert(kr); 938 } 939 940 } 941 free(buf); 942 return (0); 943 } 944 945 int 946 fetchifs(int ifindex) 947 { 948 size_t len; 949 int mib[6]; 950 char *buf, *next, *lim; 951 struct if_msghdr ifm; 952 struct kif_node *kif; 953 struct sockaddr *sa, *rti_info[RTAX_MAX]; 954 struct sockaddr_dl *sdl; 955 956 mib[0] = CTL_NET; 957 mib[1] = PF_ROUTE; 958 mib[2] = 0; 959 mib[3] = AF_INET; 960 mib[4] = NET_RT_IFLIST; 961 mib[5] = ifindex; 962 963 if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) { 964 log_warn("sysctl"); 965 return (-1); 966 } 967 if ((buf = malloc(len)) == NULL) { 968 log_warn("fetchif"); 969 return (-1); 970 } 971 if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) { 972 log_warn("sysctl"); 973 free(buf); 974 return (-1); 975 } 976 977 lim = buf + len; 978 for (next = buf; next < lim; next += ifm.ifm_msglen) { 979 memcpy(&ifm, next, sizeof(ifm)); 980 if (ifm.ifm_version != RTM_VERSION) 981 continue; 982 if (ifm.ifm_type != RTM_IFINFO) 983 continue; 984 985 sa = (struct sockaddr *)(next + sizeof(ifm)); 986 get_rtaddrs(ifm.ifm_addrs, sa, rti_info); 987 988 if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) { 989 log_warn("fetchifs"); 990 free(buf); 991 return (-1); 992 } 993 994 kif->k.ifindex = ifm.ifm_index; 995 kif->k.flags = ifm.ifm_flags; 996 kif->k.link_state = ifm.ifm_data.ifi_link_state; 997 kif->k.if_type = ifm.ifm_data.ifi_type; 998 kif->k.baudrate = ifm.ifm_data.ifi_baudrate; 999 kif->k.mtu = ifm.ifm_data.ifi_mtu; 1000 kif->k.nh_reachable = (kif->k.flags & IFF_UP) && 1001 LINK_STATE_IS_UP(ifm.ifm_data.ifi_link_state); 1002 if ((sa = rti_info[RTAX_IFP]) != NULL) 1003 if (sa->sa_family == AF_LINK) { 1004 sdl = (struct sockaddr_dl *)sa; 1005 if (sdl->sdl_nlen >= sizeof(kif->k.ifname)) 1006 memcpy(kif->k.ifname, sdl->sdl_data, 1007 sizeof(kif->k.ifname) - 1); 1008 else if (sdl->sdl_nlen > 0) 1009 memcpy(kif->k.ifname, sdl->sdl_data, 1010 sdl->sdl_nlen); 1011 /* string already terminated via calloc() */ 1012 } 1013 1014 kif_insert(kif); 1015 } 1016 free(buf); 1017 return (0); 1018 } 1019 1020 int 1021 dispatch_rtmsg(void) 1022 { 1023 char buf[RT_BUF_SIZE]; 1024 ssize_t n; 1025 char *next, *lim; 1026 struct rt_msghdr *rtm; 1027 struct if_msghdr ifm; 1028 struct sockaddr *sa, *rti_info[RTAX_MAX]; 1029 struct sockaddr_in *sa_in; 1030 struct sockaddr_rtlabel *label; 1031 struct kroute_node *kr; 1032 struct in_addr prefix, nexthop, netmask; 1033 struct iface *iface = NULL; 1034 int flags; 1035 u_short ifindex = 0; 1036 u_int8_t metric, prio; 1037 1038 if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) { 1039 if (errno == EAGAIN || errno == EINTR) 1040 return (0); 1041 log_warn("dispatch_rtmsg: read error"); 1042 return (-1); 1043 } 1044 1045 if (n == 0) { 1046 log_warnx("routing socket closed"); 1047 return (-1); 1048 } 1049 1050 lim = buf + n; 1051 for (next = buf; next < lim; next += rtm->rtm_msglen) { 1052 rtm = (struct rt_msghdr *)next; 1053 if (lim < next + sizeof(u_short) || 1054 lim < next + rtm->rtm_msglen) 1055 fatalx("dispatch_rtmsg: partial rtm in buffer"); 1056 if (rtm->rtm_version != RTM_VERSION) 1057 continue; 1058 1059 prefix.s_addr = 0; 1060 netmask.s_addr = 0; 1061 flags = F_KERNEL; 1062 nexthop.s_addr = 0; 1063 prio = 0; 1064 1065 if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE || 1066 rtm->rtm_type == RTM_DELETE) { 1067 sa = (struct sockaddr *)(next + rtm->rtm_hdrlen); 1068 get_rtaddrs(rtm->rtm_addrs, sa, rti_info); 1069 1070 if (rtm->rtm_tableid != kr_state.rdomain) 1071 continue; 1072 1073 if (rtm->rtm_pid == kr_state.pid) /* cause by us */ 1074 continue; 1075 1076 if (rtm->rtm_errno) /* failed attempts... */ 1077 continue; 1078 1079 /* Skip ARP/ND cache and broadcast routes. */ 1080 if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST)) 1081 continue; 1082 1083 prio = rtm->rtm_priority; 1084 1085 switch (sa->sa_family) { 1086 case AF_INET: 1087 prefix.s_addr = 1088 ((struct sockaddr_in *)sa)->sin_addr.s_addr; 1089 sa_in = (struct sockaddr_in *) 1090 rti_info[RTAX_NETMASK]; 1091 if (sa_in != NULL) { 1092 if (sa_in->sin_len != 0) 1093 netmask.s_addr = 1094 sa_in->sin_addr.s_addr; 1095 } else if (rtm->rtm_flags & RTF_HOST) 1096 netmask.s_addr = prefixlen2mask(32); 1097 else 1098 netmask.s_addr = 1099 prefixlen2mask(prefixlen_classful( 1100 prefix.s_addr)); 1101 if (rtm->rtm_flags & RTF_STATIC) 1102 flags |= F_STATIC; 1103 if (rtm->rtm_flags & RTF_BLACKHOLE) 1104 flags |= F_BLACKHOLE; 1105 if (rtm->rtm_flags & RTF_REJECT) 1106 flags |= F_REJECT; 1107 if (rtm->rtm_flags & RTF_DYNAMIC) 1108 flags |= F_DYNAMIC; 1109 break; 1110 default: 1111 continue; 1112 } 1113 1114 ifindex = rtm->rtm_index; 1115 if ((sa = rti_info[RTAX_GATEWAY]) != NULL) { 1116 switch (sa->sa_family) { 1117 case AF_INET: 1118 nexthop.s_addr = ((struct 1119 sockaddr_in *)sa)->sin_addr.s_addr; 1120 break; 1121 case AF_LINK: 1122 flags |= F_CONNECTED; 1123 break; 1124 } 1125 } 1126 } 1127 1128 switch (rtm->rtm_type) { 1129 case RTM_ADD: 1130 case RTM_CHANGE: 1131 if (nexthop.s_addr == 0 && !(flags & F_CONNECTED)) { 1132 log_warnx("dispatch_rtmsg no nexthop for %s/%u", 1133 inet_ntoa(prefix), 1134 mask2prefixlen(netmask.s_addr)); 1135 continue; 1136 } 1137 1138 if ((kr = kroute_find(prefix.s_addr, netmask.s_addr, 1139 prio)) != NULL) { 1140 if (kr->r.flags & F_REDISTRIBUTED) 1141 flags |= F_REDISTRIBUTED; 1142 kr->r.nexthop.s_addr = nexthop.s_addr; 1143 kr->r.flags = flags; 1144 kr->r.ifindex = ifindex; 1145 kr->r.priority = prio; 1146 1147 rtlabel_unref(kr->r.rtlabel); 1148 kr->r.rtlabel = 0; 1149 if ((label = (struct sockaddr_rtlabel *) 1150 rti_info[RTAX_LABEL]) != NULL) 1151 kr->r.rtlabel = 1152 rtlabel_name2id(label->sr_label); 1153 1154 if (kif_validate(kr->r.ifindex)) 1155 kr->r.flags &= ~F_DOWN; 1156 else 1157 kr->r.flags |= F_DOWN; 1158 1159 /* just readd, the RDE will care */ 1160 kr_redistribute(IMSG_NETWORK_ADD, &kr->r); 1161 } else { 1162 if ((kr = calloc(1, 1163 sizeof(struct kroute_node))) == NULL) { 1164 log_warn("dispatch_rtmsg"); 1165 return (-1); 1166 } 1167 1168 iface = if_find_index(rtm->rtm_index); 1169 if (iface != NULL) 1170 metric = iface->cost; 1171 else 1172 metric = DEFAULT_COST; 1173 1174 kr->r.prefix.s_addr = prefix.s_addr; 1175 kr->r.netmask.s_addr = netmask.s_addr; 1176 kr->r.nexthop.s_addr = nexthop.s_addr; 1177 kr->r.metric = metric; 1178 kr->r.flags = flags; 1179 kr->r.ifindex = ifindex; 1180 1181 if ((label = (struct sockaddr_rtlabel *) 1182 rti_info[RTAX_LABEL]) != NULL) 1183 kr->r.rtlabel = 1184 rtlabel_name2id(label->sr_label); 1185 1186 kroute_insert(kr); 1187 } 1188 break; 1189 case RTM_DELETE: 1190 if ((kr = kroute_find(prefix.s_addr, netmask.s_addr, 1191 prio)) == NULL) 1192 continue; 1193 if (!(kr->r.flags & F_KERNEL)) 1194 continue; 1195 if (kroute_remove(kr) == -1) 1196 return (-1); 1197 break; 1198 case RTM_IFINFO: 1199 memcpy(&ifm, next, sizeof(ifm)); 1200 if_change(ifm.ifm_index, ifm.ifm_flags, 1201 &ifm.ifm_data); 1202 break; 1203 case RTM_IFANNOUNCE: 1204 if_announce(next); 1205 break; 1206 default: 1207 /* ignore for now */ 1208 break; 1209 } 1210 } 1211 return (0); 1212 } 1213